Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
ACM SIGKDD Conference Tutorial, Washington, D.C., July 25, 2010 Mining Heterogeneous Information Networks Jiawei Han† Yizhou Sun† Xifeng Yan§ Philip S. Yu‡ †University of Illinois at Urbana‐Champaign § University of California at Santa Barbara ‡University of Illinois at Chicago Acknowledgements: NSF, ARL, NASA, AFOSR (MURI), Microsoft, IBM, Yahoo!, Google, HP Lab & Boeing July 12, 2010 1 Outline Motivation: Why Mining Heterogeneous Information Networks? Part I: Clustering, Ranking and Classification Clustering and Ranking in Information Networks Classification of Information Networks Part II: Data Quality and Search in Information Networks Data Cleaning and Data Validation by InfoNet Analysis Similarity Search in Information Networks Part III: Advanced Topics on Information Network Analysis Role Discovery and OLAP in Information Networks Mining Evolution and Dynamics of Information Networks Conclusions 2 Outline Motivation: Why Mining Heterogeneous Information Networks? Part I: Clustering, Ranking and Classification Clustering and Ranking in Information Networks Classification of Information Networks Part II: Data Quality and Search in Information Networks Data Cleaning and Data Validation by InfoNet Analysis Similarity Search in Information Networks Part III: Advanced Topics on Information Network Analysis Role Discovery and OLAP in Information Networks Mining Evolution and Dynamics of Information Networks Conclusions 3 What Are Information Networks? Information network: A network where each node represents an entity (e.g., actor in a social network) and each link (e.g., tie) a relationship between entities Each node/link may have attributes, labels, and weights Link may carry rich semantic information Homogeneous vs. heterogeneous networks Homogeneous networks Single object type and single link type Single model social networks (e.g., friends) WWW: a collection of linked Web pages Heterogeneous, multi‐typed networks Multiple object and link types Medical network: patients, doctors, disease, contacts, treatments Bibliographic network: publications, authors, venues 4 Ubiquitous Information Networks Graphs and substructures Chemical compounds, computer vision objects, circuits, XML Biological networks Bibliographic networks: DBLP, ArXiv, PubMed, … Social networks: Facebook >100 million active users World Wide Web (WWW): > 3 billion nodes, > 50 billion arcs Cyber‐physical networks An Internet Web Yeast protein interaction network Co-author network Social network sites 5 Homogeneous vs. Heterogeneous Networks Co-author Network Conference-Author Network 6 DBLP: An Interesting and Familiar Network DBLP: A computer science publication bibliographic database 1.4 M records (papers), 0.7 M authors, 5 K conferences, … Will this database disclose interesting knowledge about computer science research? What are the popular research fields/subfields in CS? Who are the leading researchers on DB or XQueries? How do the authors in this subfield collaborate and evolve? How many Wei Wang’s in DBLP, which paper done by which? Who is Sergy Brin’s supervisor and when? Who are very similar to Christos Faloutsos? …… All these kinds of questions, and potentially much more, can be nicely answered by the DBLP‐InfoNet How? Exploring the power of links in information networks! 7 Homo. vs. Hetero.: Differences in DB‐InfoNet Mining Homogeneous networks can often be derived from their original heterogeneous networks Coauthor networks can be derived from author‐paper‐ conference networks by projection on authors only Paper citation networks can be derived from a complete bibliographic network with papers and citations projected Heterogeneous DB‐InfoNet carries richer information than its corresponding projected homogeneous networks Typed heterogeneous InfoNet vs. non‐typed hetero. InfoNet (i.e., not distinguishing different types of nodes) Typed nodes and links imply a more structured InfoNet, and thus often lead to more informative discovery Our emphasis: Mining “structured” information networks! 8 Why Mining Heterogeneous Information Networks? Most datasets can be “organized” or “transformed” into a “structured” heterogeneous information network! Examples: DBLP, IMDB, Flickr, Google News, Wikipedia, … Structures can be progressively extracted from less organized data sets by information network analysis Information‐rich, inter‐related, organized data sets form one or a set of gigantic, interconnected, multi‐typed heterogeneous information networks Surprisingly rich knowledge can be derived from such structured heterogeneous information networks Our goal: Uncover knowledge hidden from “organized” data Exploring the power of multi‐typed, heterogeneous links Mining “structured” heterogeneous information networks! 9 Outline Motivation: Why Mining Heterogeneous Information Networks? Part I: Clustering, Ranking and Classification Clustering and Ranking in Information Networks Classification of Information Networks Part II: Data Quality and Search in Information Networks Data Cleaning and Data Validation by InfoNet Analysis Similarity Search in Information Networks Part III: Advanced Topics on Information Network Analysis Role Discovery and OLAP in Information Networks Mining Evolution and Dynamics of Information Networks Conclusions 10 Clustering and Ranking in Information Networks Integrated Clustering and Ranking of Heterogeneous Information Networks Clustering of Homogeneous Information Networks LinkClus: Clustering with link‐based similarity measure SCAN: Density‐based clustering of networks Others Spectral clustering Modularity‐based clustering Probabilistic model‐based clustering User‐Guided Clustering of Information Networks 11 Clustering and Ranking in Heterogeneous Information Networks Ranking & clustering each provides a new view over a network Ranking globally without considering clusters → dumb Ranking DB and Architecture confs. together? Clustering authors in one huge cluster without distinction? Dull to view thousands of objects (this is why PageRank!) RankClus: Integrates clustering with ranking Conditional ranking relative to clusters Uses highly ranked objects to improve clusters Qualities of clustering and ranking are mutually enhanced Y. Sun, J. Han, et al., “RankClus: Integrating Clustering with Ranking for Heterogeneous Information Network Analysis”, EDBT'09. 12 Global Ranking vs. Cluster‐Based Ranking A Toy Example: Two areas with 10 conferences and 100 authors in each area 13 RankClus: A New Framework Sub-Network Ranking Clustering 14 The RankClus Philosophy Why integrated Ranking and Clustering? Ranking and clustering can be mutually improved Ranking: Once a cluster becomes more accurate, ranking will be more reasonable for such a cluster and will be the distinguished feature of the cluster Clustering: Once ranking is more distinguished from each other, the clusters can be adjusted and get more accurate results Not every object should be treated equally in clustering! Objects preserve similarity under new measure space E.g., VLDB vs. SIGMOD 15 RankClus: Algorithm Framework Step 0. Initialization Step 1. Ranking Randomly partition target objects into K clusters Ranking for each sub‐network induced from each cluster, which serves as feature for each cluster Step 2. Generating new measure space Estimate mixture model coefficients for each target object Step 3. Adjusting cluster Step 4. Repeating Steps 1‐3 until stable 16 Focus on a Bi‐Typed Network Case Conference‐author network, links can exist between Conference (X) and author (Y) Author (Y) and author (Y) Use W to denote the links and there weights W = 17 Step 1: Ranking: Feature Extraction Two ranking strategies: Simple ranking vs. authority ranking Simple Ranking Proportional to degree counting for objects, e.g., # of publications of an author Considers only immediate neighborhood in the network Authority Ranking: Extension to HITS in weighted bi‐type network Rule 1: Highly ranked authors publish many papers in highly ranked conferences Rule 2: Highly ranked conferences attract many papers from many highly ranked authors Rule 3: The rank of an author is enhanced if he or she co‐ authors with many authors or many highly ranked authors 18 Encoding Rules in Authority Ranking Rule 1: Highly ranked authors publish many papers in highly ranked conferences Rule 2: Highly ranked conferences attract many papers from many highly ranked authors Rule 3: The rank of an author is enhanced if he or she co‐ authors with many authors or many highly ranked authors 19 Example: Authority Ranking in the 2‐Area Conference‐Author Network The rankings of authors are quite distinct from each other in the two clusters 20 Step 2: Generate New Measure Space: A Mixture Model Method Consider each target object’s links are generated under a mixture distribution of ranking from each cluster Consider ranking as a distribution: r(Y) → p(Y) Each target object xi is mapped into a K‐vector (πi,k) Parameters are estimated using the EM algorithm Maximize the log‐likelihood given all the observations of links 21 Example: 2‐D Coefficients in the 2‐Area Conference‐Author Network The conferences are well separated in the new measure space Scatter plots of two conferences and component coefficients 22 Step 3: Cluster Adjustment in New Measure Space Cluster center in new measure space Vector mean of objects in the cluster (K‐dimensional) Cluster adjustment Distance measure: 1‐Cosine similarity Assign to the cluster with the nearest center Why Better Ranking Function Derives Better Clustering? Consider the measure space generation process Highly ranked objects in a cluster play a more important role to decide a target object’s new measure Intuitively, if we can find the highly ranked objects in a cluster, equivalently, we get the right cluster 23 Step‐by‐Step Running Case Illustration Initially, ranking distributions are mixed together Two clusters of objects mixed together, but preserve similarity somehow Improved a little Two clusters are almost well separated Improved significantly Well separated Stable 24 Time Complexity: Linear to # of Links At each iteration, |E|: edges in network, m: number of target objects, K: number of clusters Ranking for sparse network Mixture model estimation ~O(mK^2) In all, linear to |E| ~O(K|E|+mK) Cluster adjustment ~O(|E|) ~O(K|E|) Note: SimRank will be at least quadratic at each iteration since it evaluates distance between every pair in the network 25 Case Study: Dataset: DBLP All the 2676 conferences and 20,000 authors with most publications, from the time period of year 1998 to year 2007 Both conference‐author relationships and co‐author relationships are used K=15 (select only 5 clusters here) 26 NetClus: Ranking & Clustering with Star Network Schema Beyond bi‐typed information network: A Star Network Schema Split a network into different layers, each representing by a net‐ cluster 27 StarNet: Schema & Net‐Cluster Star Network Schema Center type: Target type E.g., a paper, a movie, a tagging event A center object is a co‐occurrence of a bag of different types of objects, which stands for a multi‐relation among different types of objects Surrounding types: Attribute (property) types NetCluster Given a information network G, a net‐cluster C contains two pieces of information: Node set and link set as a sub‐network of G Membership indicator for each node x: P(x in C) Given a information network G, cluster number K, a clustering for G is a set of net‐clusters and for each node x, the sum of x’s probability distribution in all K net‐clusters should be 1 28 Venue Author Publish Write Research Paper Contain Term DBLP 29 StarNet of Delicious.com Web Site User Tagging Event Contain Tag Delicious.com 30 StartNet for IMDB Actor/A ctress Star in Director Direct Movie Contain Title/ Plot IMDB 31 Ranking Functions Ranking an object x of type Tx in a network G, denoted as p(x|Tx, G) Give a score to each object, according to its importance Different rules defined different ranking functions: Simple Ranking Ranking score is assigned according to the degree of an object Authority Ranking Ranking score is assigned according to the mutual enhancement by propagations of score through links “highly ranked conferences accept many good papers published by many highly ranked authors ,and highly ranked authors publish many good papers in highly ranked conferences”: 32 Ranking Function (Cont.) Priors can be added: PP(X|Tx, Gk) = (1 ‐ λP) P(X|Tx, Gk) + λP P0(X|Tx, Gk) P0(X|Tx, Gk) is the prior knowledge, usually given as a distribution, denoted by only several words λP is the parameter that we believe in prior distribution Ranking distribution Normalize ranking scores to 1, given them a probabilistic meaning Similar to the idea of PageRank 33 NetClus: Algorithm Framework Map each target object into a new low‐dimensional feature space according to current net‐clustering, and adjust the clustering further in the new measure space Step 0: Generate initial random clusters Step 1: Generate ranking‐based generative model for target objects for each net‐cluster Step 2: Calculate posterior probabilities for target objects, which serves as the new measure, and assign target objects to the nearest cluster accordingly Step 3: Repeat steps 1 and 2, until clusters do not change significantly Step 4: Calculate posterior probabilities for attribute objects in each net‐cluster 34 Generative Model for Target Objects Given a Net-cluster Each target object stands for an co‐occurrence of a bag of attribute objects Define the probability of a target object <=> define the probability of the co‐occurrence of all the associated attribute objects Generative probability P(d|Gk) for target object d in cluster Ck : where P(x |Tx , Gk) is ranking function, P(Tx |Gk) is type probability Two assumptions of independence The probabilities to visit objects of different types are independent to each other The probabilities to visit two objects within the same type are independent to each other 35 Cluster Adjustment Using posterior probabilities of target objects as new feature space Each target object => K‐dimension vector Each net‐cluster center => K‐dimension vector Average on the objects in the cluster Assign each target object into nearest cluster center (e.g., cosine similarity) A sub‐network corresponding to a new net‐cluster is then built by extracting all the target objects in that cluster and all linked attribute objects 39 Experiments: DBLP and Beyond Data Set: DBLP “all‐area” data set All conferences + “Top” 50K authors DBLP “four‐area” data set 20 conferences from DB, DM, ML, IR All authors from these conferences All papers published in these conferences Running case illustration 40 Accuracy Study: Experiments Accuracy, compared with PLSA, a pure text model, no other types of objects and links are used, use the same prior as in NetClus Accuracy of Paper Clustering Results Accuracy, compared with RankClus, a bi‐typed clustering method on only one type Accuracy of Conference Clustering Results 41 NetClus: Distinguishing Conferences AAAI 0.0022667 0.00899168 0.934024 0.0300042 0.0247133 CIKM 0.150053 0.310172 0.00723807 0.444524 0.0880127 CVPR 0.000163812 0.00763072 0.931496 0.0281342 0.032575 ECIR 3.47023e‐05 0.00712695 0.00657402 0.978391 0.00787288 ECML 0.00077477 0.110922 0.814362 0.0579426 0.015999 EDBT 0.573362 0.316033 0.00101442 0.0245591 0.0850319 ICDE 0.529522 0.376542 0.00239152 0.0151113 0.0764334 ICDM 0.000455028 0.778452 0.0566457 0.113184 0.0512633 ICML 0.000309624 0.050078 0.878757 0.0622335 0.00862134 IJCAI 0.00329816 0.0046758 0.94288 0.0303745 0.0187718 KDD 0.00574223 0.797633 0.0617351 0.067681 0.0672086 PAKDD 0.00111246 0.813473 0.0403105 0.0574755 0.0876289 PKDD 5.39434e‐05 0.760374 0.119608 0.052926 0.0670379 PODS 0.78935 0.113751 0.013939 0.00277417 0.0801858 SDM 0.000172953 0.841087 0.058316 0.0527081 0.0477156 SIGIR 0.00600399 0.00280013 0.00275237 0.977783 0.0106604 SIGMOD 0.689348 0.223122 0.0017703 0.00825455 0.0775055 VLDB 0.701899 0.207428 0.00100012 0.0116966 0.0779764 WSDM 0.00751654 0.269259 0.0260291 0.683646 0.0135497 WWW 0.0771186 0.270635 0.029307 0.451857 0.171082 42 NetClus: Database System Cluster database 0.0995511 databases 0.0708818 system 0.0678563 data 0.0214893 query 0.0133316 systems 0.0110413 queries 0.0090603 management 0.00850744 object 0.00837766 relational 0.0081175 processing 0.00745875 based 0.00736599 distributed 0.0068367 xml 0.00664958 oriented 0.00589557 design 0.00527672 web 0.00509167 information 0.0050518 model 0.00499396 efficient 0.00465707 VLDB 0.318495 SIGMOD Conf. 0.313903 ICDE 0.188746 PODS 0.107943 EDBT 0.0436849 Ranking authors in XML Surajit Chaudhuri 0.00678065 Michael Stonebraker 0.00616469 Michael J. Carey 0.00545769 C. Mohan 0.00528346 David J. DeWitt 0.00491615 Hector Garcia-Molina 0.00453497 H. V. Jagadish 0.00434289 David B. Lomet 0.00397865 Raghu Ramakrishnan 0.0039278 Philip A. Bernstein 0.00376314 Joseph M. Hellerstein 0.00372064 Jeffrey F. Naughton 0.00363698 Yannis E. Ioannidis 0.00359853 Jennifer Widom 0.00351929 Per-Ake Larson 0.00334911 Rakesh Agrawal 0.00328274 Dan Suciu 0.00309047 Michael J. Franklin 0.00304099 Umeshwar Dayal 0.00290143 Abraham Silberschatz 0.00278185 43 NetClus: StarNet‐Based Ranking and Clustering A general framework in which ranking and clustering are successfully combined to analyze infornets Ranking and clustering can mutually reinforce each other in information network analysis NetClus, an extension to RankClus that integrates ranking and clustering and generate net‐clusters in a star network with arbitrary number of types Net‐cluster, heterogeneous information sub‐networks comprised of multiple types of objects Go well beyond DBLP, and structured relational DBs Flickr: query “Raleigh” derives multiple clusters 44 iNextCube: Information Network‐Enhanced Text Cube (VLDB’09 Demo) Demo: iNextCube.cs.uiuc.edu Dimension hierarchies generated by NetClus Architecture of iNextCube Author/conferen term ranking for research area. Th research areas ca at different level All DB and IS Theory DB DM XML Distributed DB … Architecture … IR Net‐cluster Hierarchy 45 Clustering and Ranking in Information Networks Integrated Clustering and Ranking of Heterogeneous Information Networks Clustering of Homogeneous Information Networks LinkClus: Clustering with link‐based similarity measure SCAN: Density‐based clustering of networks Others Spectral clustering Modularity‐based clustering Probabilistic model‐based clustering User‐Guided Clustering of Information Networks 46 Link‐Based Clustering: Why Useful? Authors Tom Proceedings sigmod03 sigmod04 Mike Cathy John sigmod sigmod05 vldb03 vldb04 vldb vldb05 aaai04 Mary Conferences aaai05 aaai Questions: Q1: How to cluster each type of objects? Q2: How to define similarity between each type of objects? 47 SimRank: Link‐Based Similarities Two objects are similar if linked with the same or similar objects Jeh & Widom, 2002 ‐ SimRank Tom sigmod03 sigmod04 Mary Tom Mike Cathy John sigmod Similarity between two objects a and b, S(a, b) = the average similarity between objects linked with a and those with b: sigmod where I(v) is the set of in‐neighbors of the vertex v. sigmod05 sigmod03 sigmod04 sigmod05 But: It is expensive to compute: vldb03 vldb04 vldb05 vldb For a dataset of N objects and M links, it takes O(N2) space and O(M2) time to compute all similarities. 48 Observation 1: Hierarchical Structures Hierarchical structures often exist naturally among objects (e.g., taxonomy of animals) Relationships between articles and words (Chakrabarti, Papadimitriou, Modha, Faloutsos, 2004) A hierarchical structure of products in Walmart grocery electronics TV DVD apparel Articles All camera Words 49 Observation 2: Distribution of Similarity portion of entries 0.4 Distribution of SimRank similarities among DBLP authors 0.3 0.2 0.1 0.24 0.22 0.2 0.18 0.16 0.14 0.12 0.1 0.08 0.06 0.04 0.02 0 0 similarity value Power law distribution exists in similarities 56% of similarity entries are in [0.005, 0.015] 1.4% of similarity entries are larger than 0.1 Our goal: Design a data structure that stores the significant similarities and compresses insignificant ones 50 Our Data Structure: SimTree Each non-leaf node represents a group of similar lower-level nodes Similarity between two sibling nodes n1 and n2 0.2 n1 Adjustment ratio for node n7 0.8 0.9 n7 n4 0.3 0.8 n2 0.9 0.9 n5 n6 n8 n3 1.0 n9 Each leaf node simp(n7,n8) = s(n7,n4) x s(n4,n5) x s(n5,n8) represents an object Path-based node similarity Similarity between two nodes is the average similarity between objects linked with them in other SimTrees Average similarity between x and all other nodes Adjustment ratio for x = Average similarity between x’s parent and all other nodes 51 LinkClus: SimTree‐Based Hierarchical Clustering Initialize a SimTree for objects of each type Repeat For each SimTree, update the similarities between its nodes using similarities in other SimTrees Similarity between two nodes a and b is the average similarity between objects linked with them Adjust the structure of each SimTree Assign each node to the parent node that it is most similar to 52 Initialization of SimTrees Finding tight groups Frequent pattern mining Reduced to The tightness of a group of nodes is the support of a frequent pattern g1 n2 g2 n1 n3 n4 Transactions 1 2 3 4 5 6 7 8 9 {n1} {n1, n2} {n2} {n1, n2} {n1, n2} {n2, n3, n4} {n4} {n3, n4} {n3, n4} Initializing a tree: Start from leaf nodes (level‐0) At each level l, find non‐overlapping groups of similar nodes with frequent pattern mining 53 Complexity: LinkClus vs. SimRank After initialization, iteratively (1) for each SimTree update the similarities between its nodes using similarities in other SimTrees, and (2) Adjust the structure of each SimTree Computational complexity: For two types of objects, N in each, and M linkages between them Time Space Updating similarities O(M(logN)2) O(M+N) Adjusting tree structures O(N) O(N) LinkClus O(M(logN)2) O(M+N) SimRank O(M2) O(N2) 54 Performance Comparison: Experiment Setup DBLP dataset: 4170 most productive authors, and 154 well-known conferences with most proceedings Manually labeled research areas of 400 most productive authors according to their home pages (or publications) Manually labeled areas of 154 conferences according to their call for papers Approaches Compared: SimRank (Jeh & Widom, KDD 2002) Computing pair-wise similarities SimRank with FingerPrints (F-SimRank) Fogaras & R´acz, WWW 2005 pre-computes a large sample of random paths from each object and uses samples of two objects to estimate SimRank similarity ReCom (Wang et al. SIGIR 2003) Iteratively clustering objects using cluster labels of linked objects 55 DBLP Data Set: Accuracy and Computation Time Conferences Authors 0.8 1 0.7 0.3 LinkClus SimRank ReCom F-SimRank 0.2 #iteration 13 11 9 7 5 3 1 19 17 15 13 11 9 7 5 3 0.1 19 0.8 0.4 17 LinkClus SimRank ReCom F-SimRank 0.85 0.5 15 accuracy 0.6 0.9 1 accuracy 0.95 #iteration Approaches Accr‐Author Accr‐Conf average time LinkClus 0.957 0.723 76.7 SimRank 0.958 0.760 1020 ReCom 0.907 0.457 43.1 F‐SimRank 0.908 0.583 83.6 56 Email Dataset: Accuracy and Time F. Nielsen. Email dataset. http://www.imm.dtu.dk/∼rem/data/Email‐1431.zip 370 emails on conferences, 272 on jobs, and 789 spam emails Why is LinkClus even better than SimRank in accuracy? Noise filtering due to frequent pattern‐based preprocessing Approach Accuracy Total time (sec) LinkClus 0.8026 1579.6 SimRank 0.7965 39160 ReCom 0.5711 74.6 F‐SimRank 0.3688 479.7 CLARANS 0.4768 8.55 57 Clustering and Ranking in Information Networks Integrated Clustering and Ranking of Heterogeneous Information Networks Clustering of Homogeneous Information Networks LinkClus: Clustering with link‐based similarity measure SCAN: Density‐based clustering of networks Others Spectral clustering Modularity‐based clustering Probabilistic model‐based clustering User‐Guided Clustering of Information Networks 58 SCAN: Density‐Based Network Clustering Networks made up of the mutual relationships of data elements usually have an underlying structure. Clustering: A structure discovery problem Given simply information of who associates with whom, could one identify clusters of individuals with common interests or special relationships (families, cliques, terrorist cells)? Questions to be answered: How many clusters? What size should they be? What is the best partitioning? Should some points be segregated? Scan: An interesting density‐based algorithm X. Xu, N. Yuruk, Z. Feng, and T. A. J. Schweiger, “SCAN: A Structural Clustering Algorithm for Networks”, Proc. 2007 ACM SIGKDD Int. Conf. Knowledge Discovery in Databases (KDD'07), San Jose, CA, Aug. 2007 59 Social Network and Its Clustering Problem Individuals in a tight social group, or clique, know many of the same people, regardless of the size of the group. Individuals who are hubs know many people in different groups but belong to no single group. Politicians, for example bridge multiple groups. Individuals who are outliers reside at the margins of society. Hermits, for example, know few people and belong to no group. 60 Structure Similarity Define Γ(v) as immediate neighbor of a vertex v. The desired features tend to be captured by a measure σ(u, v) as Structural Similarity | Γ(v) I Γ( w) | σ (v, w) = | Γ(v) || Γ( w) | v Structural similarity is large for members of a clique and small for hubs and outliers. 61 Structural Connectivity [1] N ε (v ) = {w ∈ Γ(v) | σ (v, w) ≥ ε } ε‐Neighborhood: Core: Direct structure reachable: COREε , μ (v) ⇔| N ε (v) |≥ μ DirRECH ε , μ (v, w) ⇔ CORE ε , μ (v ) ∧ w ∈ N ε (v ) Structure reachable: transitive closure of direct structure reachability Structure connected: CONNECTε , μ (v, w) ⇔ ∃u ∈ V : RECH ε , μ (u , v ) ∧ RECH ε , μ (u , w) [1] M. Ester, H. P. Kriegel, J. Sander, & X. Xu (KDD'96) “A DensityBased Algorithm for Discovering Clusters in Large Spatial Databases 62 Structure‐Connected Clusters Structure‐connected cluster C Connectivity: Maximality: ∀v, w ∈ C : CONNECT ε , μ (v, w) ∀v, w ∈ V : v ∈ C ∧ REACH ε , μ (v, w) ⇒ w ∈ C Hubs: Not belong to any cluster Bridge to many clusters Outliers: Not belong to any cluster Connect to less clusters hub outlier 63 Algorithm 2 3 μ=2 ε = 0.7 5 1 4 0.67 8 7 0.82 12 0.75 6 11 0 10 9 13 64 Algorithm 2 3 μ=2 ε = 0.7 5 0.51 0.68 7 6 11 8 1 4 0.51 0 12 10 9 13 65 Running Time Running time: O(|E|) For sparse networks: O(|V|) [2] A. Clauset, M. E. J. Newman, & C. Moore, Phys. Rev. E 70, 066111 (2004). 66 Clustering and Ranking in Information Networks Integrated Clustering and Ranking of Heterogeneous Information Networks Clustering of Homogeneous Information Networks LinkClus: Clustering with link‐based similarity measure SCAN: Density‐based clustering of networks Others Spectral clustering Modularity‐based clustering Probabilistic model‐based clustering User‐Guided Clustering of Information Networks 67 Spectral Clustering Spectral clustering: Find the best cut that partitions the network Different criteria to decide “best” Min cut, ratio cut, normalized cut, Min‐Max cut Using Min cut as an example [Wu et al. 1993] Assign each node i an indicator variable Represent the cut size using indicator vector and adjacency matrix Cutsize = Minimize the objective function through solving eigenvalue system Relax the discrete value of q to continuous value Map continuous value of q into discrete ones to get cluster labels Use second smallest eigenvector for q 68 Modularity‐Based Clustering Modularity‐based clustering Find the best clustering that maximizes the modularity function Q‐function [Newman et al., 2004] Let eij be a half of the fraction of edges between group i and group j eii is the fraction of edges within group i Let ai be the fraction of all ends of edges attached to vertices in group I Q is then defined as sum of difference between within‐group edges and expected within‐group edges Minimize Q One possible solution: hierarchically merge clusters resulting in greatest increase in Q function [Newman et al., 2004] 69 Probabilistic Model‐Based Clustering Probabilistic model‐based clustering Build generative models for links based on hidden cluster labels Maximize the log‐likelihood of all the links to derive the hidden cluster membership An example: Airoldi et al., Mixed membership stochastic block models, 2008 Define a group interaction probability matrix B (KXK) B(g,h) denotes the probability of link generation between group g and group h Generative model for a link For each node, draw a membership probability vector form a Dirichlet prior For each paper of nodes, draw cluster labels according to their membership probability (supposing g and h), then decide whether to have a link according to probability B(g, h) Derive the hidden cluster label by maximize the likelihood given B and the prior 70 Clustering and Ranking in Information Networks Integrated Clustering and Ranking of Heterogeneous Information Networks Clustering of Homogeneous Information Networks LinkClus: Clustering with link‐based similarity measure SCAN: Density‐based clustering of networks Others Spectral clustering Modularity‐based clustering Probabilistic model‐based clustering User‐Guided Clustering of Information Networks 71 User‐Guided Clustering in DB InfoNet Open-course Course Work-In Professor person name course course-id group office semester name position instructor area Advise Group professor name student area degree User hint Target of clustering Publish Publication author title title year conf Register student Student course name office semester position unit grade User usually has a goal of clustering, e.g., clustering students by research area User specifies his clustering goal to a DB‐InfoNet cluster: CrossClus 72 Classification vs. User‐Guided Clustering User hint User‐specified feature (in the form of attribute) is used as a hint, not class labels The attribute may contain too many or too few distinct values All tuples for clustering E.g., a user may want to cluster students into 20 clusters instead of 3 Additional features need to be included in cluster analysis 73 User‐Guided Clustering vs. Semi‐supervised Clustering Semi‐supervised clustering [Wagstaff, et al’ 01, Xing, et al.’02] User provides a training set consisting of “similar” and “dissimilar” pairs of objects User‐guided clustering User specifies an attribute as a hint, and more relevant features are found for clustering Semi-supervised clustering User-guided clustering x All tuples for clustering All tuples for clustering 74 Why Not Typical Semi‐Supervised Clustering? Why not do typical semi‐supervise clustering? Much information (in multiple relations) is needed to judge whether two tuples are similar A user may not be able to provide a good training set It is much easier for a user to specify an attribute as a hint, such as a student’s research area Tom Smith Jane Chang SC1211 BI205 TA RA Tuples to be compared User hint 75 CrossClus: An Overview CrossClus: Framework Search for good multi‐relational features for clustering Measure similarity between features based on how they cluster objects into groups User guidance + heuristic search for finding pertinent features Clustering based on a k‐medoids‐based algorithm CrossClus: Major advantages User guidance, even in a very simple form, plays an important role in multi‐relational clustering CrossClus finds pertinent features by computing similarities between features 76 Selection of Multi‐Relational Features A multi‐relational feature is defined by: A join path. E.g., Student → Register → OpenCourse → Course An attribute. E.g., Course.area (For numerical feature) an aggregation operator. E.g., sum or average Categorical Feature f = [Student → Register → OpenCourse → Course, Course.area, null] areas of courses of each student Tuple t1 t2 t3 t4 t5 Areas of courses Values of feature f Feature f Tuple DB AI TH 5 0 1 5 3 5 3 5 0 3 0 7 4 5 4 t1 t2 t3 t4 t5 DB AI TH 0.5 0 0.1 0.5 0.3 0.5 0.3 0.5 0 0.3 0 0.7 0.4 0.5 0.4 f(t1) f(t2) f(t3) f(t4) DB AI TH f(t5) Numerical Feature, e.g., average grades of students h = [Student → Register, Register.grade, average] E.g. h(t1) = 3.5 77 Similarity Between Features Vf Values of Feature f and g Feature f (course) DB t1 t2 0.5 0 AI 0.5 0.3 TH 0 0.7 Feature g (group) Info sys Cog sci 1 0 0 0 1 0.9 0.5-0.6 0.8 0.4-0.5 Theory 0.7 0.3-0.4 0.6 0.5 0 0.4 0.3 0.2 0.1 0 1 0.1 0.5 0.4 0 0.5 0.5 0.5 0 0.5 0.5 0 0.5 t5 0.3 0.3 0.4 0.5 0.5 0 Similarity between two features – cosine similarity of two vectors V ⋅V Sim( f , g ) = f g V V f g 0-0.1 3 2 S4 S3 1 S2 t4 0.1-0.2 4 S5 t3 0.2-0.3 5 S1 Vg 1 0.9 0.8 0.5-0.6 0.7 0.4-0.5 0.6 0.3-0.4 0.5 0.2-0.3 0.4 1 0.3 0.2 0.1-0.2 0-0.1 2 0.1 3 0 S1 4 S2 S3 S4 5 S5 78 Similarity between Categorical & Numerical Features V ⋅V h = 2 ∑ ∑ sim h (t i , t j )⋅ sim f (ti , t j ) N f i =1 j < i N l = 2∑ ∑ i =1 k =1 N l ⎛ ⎞ ⎛ ⎞ f (ti ). p k (1 − h (t i ))⎜⎜ ∑ f (t j ). p k ⎟⎟ + 2 ∑ ∑ f (t i ). p k ⎜⎜ ∑ h (t j )⋅ f (t j ). p k ⎟⎟ i =1 k =1 ⎝ j <i ⎠ ⎝ j <i ⎠ Only depend on ti Feature f DB AI TH Objects (ordered by h) Feature h 2.7 2.9 3.1 3.3 3.5 3.7 3.9 Depend on all tj with j<i Parts depending on ti Parts depending on all ti with j<i 79 Searching for Pertinent Features Different features convey different aspects of information Research area Research group area Demographic info GPA Conferences of papers Permanent address GRE score Nationality Number of papers Advisor Academic Performances Features conveying same aspect of information usually cluster objects in more similar ways research group areas vs. conferences of publications Given user specified feature Find pertinent features by computing feature similarity 80 Heuristic Search for Pertinent Features Overall procedure Course Professor person name course course-id group office semester name position instructor area 1.Start from the user‐ specified feature Group 2. Search in neighborhood of name existing pertinent features 3. Expand search range gradually area 2 Advise student degree author 1 title Publication title year conf Register student Student Target of clustering Publish professor User hint Open-course Work-In name office position course semester unit grade Tuple ID propagation [Yin, et al.’04] is used to create multi-relational features IDs of target tuples can be propagated along any join path, from which we can find tuples joinable with each target tuple 81 Clustering with Multi‐Relational Feature Given a set of L pertinent features f1, …, fL, similarity between two objects L sim (t1 , t 2 ) = ∑ sim f i (t1 , t 2 ) ⋅ f i .weight i =1 Weight of a feature is determined in feature search by its similarity with other pertinent features For clustering, we use CLARANS, a scalable k‐medoids [Ng & Han’94] algorithm 82 Experiments: Compare CrossClus with Existing Methods Baseline: Only use the user specified feature PROCLUS [Aggarwal, et al. 99]: a state‐of‐the‐art subspace clustering algorithm Use a subset of features for each cluster We convert relational database to a table by propositionalization User‐specified feature is forced to be used in every cluster RDBC [Kirsten and Wrobel’00] A representative ILP clustering algorithm Use neighbor information of objects for clustering User‐specified feature is forced to be used 83 Measuring Clustering Accuracy To verify that CrossClus captures user’s clustering goal, we define “accuracy” of clustering Given a clustering task Manually find all features that contain information directly related to the clustering task – standard feature set E.g., Clustering students by research areas Standard feature set: research group, group areas, advisors, conferences of publications, course areas Accuracy of clustering result: how similar it is to the clustering generated by standard feature set ∑ deg (C ⊂ C ') = n i =1 ( max 1≤ j ≤ n ' ci ∩ c ' j ∑ n i =1 ) ci deg (C ⊂ C ') + deg (C ' ⊂ C ) sim (C , C ') = 2 84 Clustering Professors: CS Dept Dataset Clustering Accuracy - CS Dept 1 0.8 CrossClus K-Medoids CrossClus K-Means CrossClus Agglm Baseline PROCLUS RDBC 0.6 0.4 0.2 0 Group Course Group+Course (Theory): J. Erickson, S. Har‐Peled, L. Pitt, E. Ramos, D. Roth, M. Viswanathan (Graphics): J. Hart, M. Garland, Y. Yu (Database): K. Chang, A. Doan, J. Han, M. Winslett, C. Zhai (Numerical computing): M. Heath, T. Kerkhoven, E. de Sturler (Networking & QoS): R. Kravets, M. Caccamo, J. Hou, L. Sha (Artificial Intelligence): G. Dejong, M. Harandi, J. Ponce, L. Rendell (Architecture): D. Padua, J. Torrellas, C. Zilles, S. Adve, M. Snir, D. Reed, V. Adve (Operating Systems): D. Mickunas, R. Campbell, Y. Zhou 85 DBLP Dataset Clustering Accurarcy - DBLP 1 0.9 0.8 0.7 CrossClus K-Medoids CrossClus K-Means CrossClus Agglm Baseline PROCLUS RDBC 0.6 0.5 0.4 0.3 0.2 0.1 th re e A ll th or +C oa u W or d Co au th or or d Co nf + Co nf + W or Co au th or d W Co nf 0 86 Outline Motivation: Why Mining Heterogeneous Information Networks? Part I: Clustering, Ranking and Classification Clustering and Ranking in Information Networks Classification of Information Networks Part II: Data Quality and Search in Information Networks Data Cleaning and Data Validation by InfoNet Analysis Similarity Search in Information Networks Part III: Advanced Topics on Information Network Analysis Role Discovery and OLAP in Information Networks Mining Evolution and Dynamics of Information Networks Conclusions 87 Classification of Information Networks Classification of Heterogeneous Information Networks: Graph‐regularization‐Based Method (GNetMine) Multi‐Relational‐Mining‐Based Method (CrossMine) Statistical Relational Learning‐Based Method (SRL) Classification of Homogeneous Information Networks 88 Why Classifying Heterogeneous InfoNet? Sometimes, we do have prior knowledge for part of the nodes/objects! Input: Heterogeneous information network structure + class labels for some objects/nodes Goal: Classify the heterogeneous networked data into classes, each of which is composed of multi‐typed data objects sharing a common topic. Natural generalization of classification on homogeneous networked data Find out the terrorists, their emails and frequently used words! Class: terrorism Email network + several suspicious users/words/emails Military network + which military camp several soldiers/commanders belong to …… Classifier Find out the soldiers and commanders belonging to that camp! Class: military camp …… 89 Classification: Knowledge Propagation 90 GNetMine: Methodology Classification of networked data can be essentially viewed as a process of knowledge propagation, where information is propagated from labeled objects to unlabeled ones through links until a stationary state is achieved. A novel graph‐based regularization framework to address the classification problem on heterogeneous information networks. Respect the link type differences by preserving consistency over each relation graph corresponding to each type of links separately Mathematical intuition: Consistency assumption The confidence (f)of two objects (xip and xjq) belonging to class k should be similar if xip ↔ xjq (Rij,pq > 0) f should be similar to the given ground truth 91 GNetMine: Graph‐Based Regularization Minimize the objective function User preference: how much do you J ( f1( k ) ,..., f m( k ) ) nj value this relationship / ground truth? 1 1 (k ) f ip − f jq( k ) ) 2 = ∑ λij ∑∑ Rij , pq ( Dij , pp D ji ,qq i , j =1 p =1 q =1 ni m m + ∑ α i ( f i ( k ) − y i( k ) )T ( f i ( k ) − y i( k ) ) i =1 Smoothness constraints: objects linked together should share similar estimations of confidence belonging to class k Normalization term applied to each type of link separately: reduce the impact of popularity of nodes Confidence estimation on labeled data and their pre‐given labels should be similar 92 Experiments on DBLP Class: Four research areas (communities) Database, data mining, AI, information retrieval Four types of objects Paper (14376), Conf. (20), Author (14475), Term (8920) Three types of relations Paper‐conf., paper‐author, paper‐term Algorithms for comparison Learning with Local and Global Consistency (LLGC) [Zhou et al. NIPS 2003] – also the homogeneous version of our method Weighted‐vote Relational Neighbor classifier (wvRN) [Macskassy et al. JMLR 2007] Network‐only Link‐based Classification (nLB) [Lu et al. ICML 2003, Macskassy et al. JMLR 2007] 93 Classification Accuracy: Labeling a Very Small Portion of Authors and Papers (a%, p%) (0.1%, 0.1%) (0.2%, 0.2%) (0.3%, 0.3%) (0.4%, 0.4%) (0.5%, 0.5%) nLB A‐A 25.4 28.3 28.4 30.7 29.8 A‐C‐P‐T 26.0 26.0 27.4 26.7 27.3 A‐A 40.8 46.0 48.6 46.3 49.0 wvRN A‐C‐P‐T 34.1 41.2 42.5 45.6 51.4 A‐A 41.4 44.7 48.8 48.7 50.6 LLGC A‐C‐P‐T 61.3 62.2 65.7 66.0 68.9 GNetMine A‐C‐P‐T 82.9 83.4 86.7 87.2 87.5 Comparison of classification accuracy on authors (%) (a%, p%) (0.1%, 0.1%) (0.2%, 0.2%) (0.3%, 0.3%) (0.4%, 0.4%) (0.5%, 0.5%) nLB P‐P 49.8 73.1 77.9 79.1 80.7 A‐C‐P‐T 31.5 40.3 35.4 38.6 39.3 P‐P 62.0 71.7 77.9 78.1 77.9 wvRN A‐C‐P‐T 42.0 49.7 54.3 54.4 53.5 P‐P 67.2 72.8 76.8 77.9 79.0 LLGC A‐C‐P‐T 62.7 65.5 66.6 70.5 73.5 GNetMine A‐C‐P‐T 79.2 83.5 83.2 83.7 84.1 Comparison of classification accuracy on papers (%) (a%, p%) (0.1%, 0.1%) (0.2%, 0.2%) (0.3%, 0.3%) (0.4%, 0.4%) (0.5%, 0.5%) nLB A‐C‐P‐T 25.5 22.5 25.0 25.0 25.0 wvRN A‐C‐P‐T 43.5 56.0 59.0 57.0 68.0 LLGC A‐C‐P‐T 79.0 83.5 87.0 86.5 90.0 GNetMine A‐C‐P‐T 81.0 85.0 87.0 89.5 94.0 Comparison of classification accuracy on conferences(%) 94 Knowledge Propagation: List Objects with the Highest Confidence Measure Belonging to Each Class No. Database Data Mining Artificial Intelligence Information Retrieval 1 data mining learning retrieval 2 database data knowledge information 3 query clustering Reinforcement web 4 5 system xml learning classification reasoning model search document Top‐5 terms related to each area No. 1 2 3 4 5 Database Surajit Chaudhuri H. V. Jagadish Michael J. Carey Michael Stonebraker C. Mohan Data Mining Jiawei Han Philip S. Yu Christos Faloutsos Wei Wang Shusaku Tsumoto Artificial Intelligence Sridhar Mahadevan Takeo Kanade Andrew W. Moore Satinder P. Singh Thomas S. Huang Information Retrieval W. Bruce Croft Iadh Ounis Mark Sanderson ChengXiang Zhai Gerard Salton Top‐5 authors concentrated in each area No. 1 2 3 4 5 Database VLDB SIGMOD PODS ICDE EDBT Data Mining KDD SDM PAKDD ICDM PKDD Artificial Intelligence IJCAI AAAI CVPR ICML ECML Information Retrieval SIGIR ECIR WWW WSDM CIKM Top‐5 conferences concentrated in each area 95 Classification of Information Networks Classification of Heterogeneous Information Networks: Graph‐regularization‐Based Method (GNetMine) Multi‐Relational‐Mining‐Based Method (CrossMine) Statistical Relational Learning‐Based Method (SRL) Classification of Homogeneous Information Networks 96 Multi‐Relation to Flat Relation Mining? Folding multiple relations into a single “flat” one for mining? Doctor Patient Contact flatten Cannot be a solution due to problems: Lose information of linkages and relationships, no semantics preservation Cannot utilize information of database structures or schemas (e.g., E‐R modeling) 97 One Approach: Inductive Logic Programming (ILP) Find a hypothesis that is consistent with background knowledge (training data) FOIL, Golem, Progol, TILDE, … Background knowledge Relations (predicates), Tuples (ground facts) Inductive Logic Programming (ILP) Hypothesis: The hypothesis is usually a set of rules, which can predict certain attributes in certain relations Daughter(X, Y) ← female(X), parent(Y, X) Training examples Daughter(mary, ann) Daughter(eve, tom) Daughter(tom, ann) Daughter(eve, ann) Background knowledge + Parent(ann, mary) + Parent(ann, tom) – Parent(tom, eve) – Parent(tom, ian) Female(ann) Female(mary) Female(eve) 98 Inductive Logic Programming Approach to Multi‐ Relation Classification ILP Approached to Multi‐Relation Classification Top‐down Approaches (e.g., FOIL) while(enough examples left) generate a rule remove examples satisfying this rule Bottom‐up Approaches (e.g., Golem) Use each example as a rule Generalize rules by merging rules Decision Tree Approaches (e.g., TILDE) ILP Approach: Pros and Cons Advantages: Expressive and powerful, and rules are understandable Disadvantages: Inefficient for databases with complex schemas, and inappropriate for continuous attributes 99 FOIL: First‐Order Inductive Learner (Rule Generation) Find a set of rules consistent with training data A top‐down, sequential covering learner Build each rule by heuristics Foil gain – a special type of information gain Examples covered by Rule 2 Examples covered Examples covered by Rule 1 by Rule 3 All examples A3=1&&A1=2 A3=1&&A1=2 &&A8=5 A3=1 Positive examples Negative examples To generate a rule while(true) find the best predicate p if foil‐gain(p)>threshold then add p to current rule else break 100 Find the Best Predicate: Predicate Evaluation All predicates in a relation can be evaluated based on propagated IDs Use foil‐gain to evaluate predicates Suppose current rule is r. For a predicate p, foil‐gain(p) = ⎤ ⎡ P(r ) P(r + p ) P(r + p )× ⎢− log + log ⎥ ( ) ( ) ( ) ( ) P r + N r P r p N r p + + + ⎦ ⎣ Categorical Attributes Compute foil‐gain directly Numerical Attributes Discretize with every possible value 101 Loan Applications: Backend Database Loan Target relation: Each tuple has a class label, indicating whether a loan is paid on time. loan-id account-id date Account District account-id district-id district-id dist-name frequency date amount duration payment Transaction trans-id Card region card-id #people disp-id #lt-500 type #lt-2000 issue-date #lt-10000 #gt-10000 account-id date Order order-id account-id bank-to account-to amount type #city Disposition type disp-id operation account-id amount client-id balance symbol ratio-urban avg-salary unemploy95 unemploy96 den-enter Client client-id #crime95 #crime96 birth-date gender district-id How to make decisions to loan applications? 102 CrossMine: An Effective Multi‐relational Classifier Methodology Tuple‐ID propagation: an efficient and flexible method for virtually joining relations Confine the rule search process in promising directions Look‐one‐ahead: a more powerful search strategy Negative tuple sampling: improve efficiency while maintaining accuracy 103 Tuple ID Propagation Loan ID Applicant #1 Applicant #3 Amount Duration Decision 1 124 1000 12 Yes 2 124 4000 12 Yes 3 108 10000 24 No 4 45 12000 36 No Account ID Applicant #2 Account ID Frequency Open date Propagated ID Labels 124 monthly 02/27/93 1, 2 2+, 0– 108 weekly 09/23/97 3 0+, 1– 45 monthly 12/09/96 4 0+, 1– 67 weekly 01/01/97 Null 0+, 0– Possible predicates: •Frequency=‘monthly’: 2 +, 1 – •Open date < 01/01/95: 2 +, 0 – Applicant #4 Propagate tuple IDs of target relation to non‐target relations Virtually join relations to avoid the high cost of physical joins 104 Tuple ID Propagation (Idea Outlined) Efficient Only propagate the tuple IDs Time and space usage is low Flexible Can propagate IDs among non‐target relations Many sets of IDs can be kept on one relation, which are propagated from different join paths Target Relation R1 R2 R3 105 Rule Generation: Example Target relation Loan loan‐id account‐id date Rule Generation Start at the target relation Repeat Search in all active relations Search in all relations joinable to active relations Add the best predicate to the current rule Set the involved relation to active Until The best predicate does not have enough gain Current rule is too long amount duration payment Account District account‐id district‐id district‐id dist‐name frequency Card region date card‐id #people disp‐id #lt‐500 type #lt‐2000 issue‐date #lt‐10000 #gt‐10000 First predicate Transaction trans‐id account‐id date Order order‐id account‐id bank‐to account‐to amount type #city Disposition type disp‐id operation account‐id amount client‐id balance symbol ratio‐urban avg‐salary unemploy95 unemploy96 Second predicate den‐enter Client client‐id #crime95 #crime96 birth‐date gender district‐id 106 Look‐one‐ahead in Rule Generation Two types of relations: Entity and Relationship Often cannot find useful predicates on relations of relationship No good predicate Target Relation Solution of CrossMine: When propagating IDs to a relation of relationship, propagate one more step to next relation of entity 107 Negative Tuple Sampling Each time a rule is generated, covered positive examples are removed After generating many rules, there are much less positive examples than negative ones Cannot build good rules (low support) Still time consuming (large number of negative examples) Solution: Sampling on negative examples Improve efficiency without affecting rule quality + – – –+–++ – + – + +– + – +– – – – + – ++ – – + + –+ + – – ––– – – –– +– – – –– – – ++ –+ + – 108 Real Dataset PKDD Cup 99 dataset – Loan Application Accuracy Time (per fold) FOIL 74.0% 3338 sec TILDE 81.3% 2429 sec CrossMine 90.7% 15.3 sec Mutagenesis dataset (4 relations): Only 4 relations, so TILDE does a good job, though slow Accuracy Time (per fold) FOIL 79.7% 1.65 sec TILDE 89.4% 25.6 sec CrossMine 87.7% 0.83 sec 109 Classification of Information Networks Classification of Heterogeneous Information Networks: Graph‐regularization‐Based Method (GNetMine) Multi‐Relational‐Mining‐Based Method (CrossMine) Statistical Relational Learning‐Based Method (SRL) Classification of Homogeneous Information Networks 110 Probabilistic Relational Models in Statistical Relational Learning Goal: model distribution of data in relational databases Treat both entities and relations as classes Intuition: objects are no longer independent with each other Build statistical networks according to the dependency relationships between attributes of different classes A Probabilistic Relational Models (PRM) consists of Relational schema (from databases) Dependency structure (between attributes) Local probability model (conditional probability distribution) Three major methods of probabilistic relational models Relational Bayesian Network (RBN, Lise Getoor et al.) Relational Markov Networks (RMN, Ben Taskar et al.) Relational Dependency Networks (RDN, Jennifer Neville et al.) 111 Relational Bayesian Networks (RBN) Extend Bayesian network to consider entities, properties and relationships in a DB scenario Three different uncertainties Attribute uncertainty Model conditional probability for an attribute given its parent variables Structural uncertainty Model conditional probability for a reference or link existence given its parent variables Class uncertainty Refine the conditional probability by considering sub‐ classes or hierarchy of classes 112 Rel. Markov Networks & Rel. Dependency Networks Similar ideas to Relational Bayesian Networks Relational Markov Networks Extend from Markov Network Undirected link to model dependency relation instead of directed links as in Bayesian networks Relational Dependency Networks Extend from Dependency Network Undirected link to model dependency relation Use pseudo‐likelihood instead of exact likelihood Efficient in learning 113 Classification of Information Networks Classification of Heterogeneous Information Networks: Graph‐regularization‐Based Method (GNetMine) Multi‐Relational‐Mining‐Based Method (CrossMine) Statistical Relational Learning‐Based Method (SRL) Classification of Homogeneous Information Networks 114 Transductive Learning in the Graph Problem: for a set of nodes in the graph, the class labels are given for partial of the nodes, the task is to learn the labels of the unlabeled nodes Methods Label propagation algorithm [Zhu et al. 2002, Zhou et al. 2004, Szummer et al. 2001] Iteratively propagate labels to its neighbors, according to the transition probability defined by the network Graph regularization‐based algorithm [Zhou et al. 2004] Intuition: trade off between (1) consistency with the labeling data and (2) consistency between linked objects An quadratic optimization problem 115 Outline Motivation: Why Mining Heterogeneous Information Networks? Part I: Clustering, Ranking and Classification Clustering and Ranking in Information Networks Classification of Information Networks Part II: Data Quality and Search in Information Networks Data Cleaning and Data Validation by InfoNet Analysis Similarity Search in Information Networks Part III: Advanced Topics on Information Network Analysis Role Discovery and OLAP in Information Networks Mining Evolution and Dynamics of Information Networks Conclusions 116 Data Cleaning by Link Analysis Object reconciliation vs. object distinction as data cleaning tasks Link analysis may take advantages of redundancy and make facilitate entity cross‐checking and validation Object distinction: Different people/objects do share names In AllMusic.com, 72 songs and 3 albums named “Forgotten” or “The Forgotten” In DBLP, 141 papers are written by at least 14 “Wei Wang” New challenges of object distinction: Textual similarity cannot be used Distinct: Object distinction by information network analysis X. Yin, J. Han, and P. S. Yu, “Object Distinction: Distinguishing Objects with Identical Names by Link Analysis”, ICDE'07 117 Entity Distinction: The “Wei Wang” Challenge in DBLP (2) (1) Wei Wang, Jiong Yang, Richard Muntz VLDB Haixun Wang, Wei Wang, Jiong Yang, Philip S. Yu Jiong Yang, Hwanjo Yu, Wei Wang, Jiawei Han 1997 SIGMOD CSB VLDB 2004 Hongjun Lu, Yidong Yuan, Wei Wang, Xuemin Lin ICDE 2005 ADMA 2005 2003 Wei Wang, Xuemin Lin Jiong Yang, Jinze Liu, Wei Wang Jinze Liu, Wei Wang 2002 Wei Wang, Haifeng Jiang, Hongjun Lu, Jeffrey Yu ICDM KDD 2004 Jian Pei, Jiawei Han, Hongjun Lu, et al. ICDM 2001 2004 (4) Jian Pei, Daxin Jiang, Aidong Zhang ICDE 2005 Aidong Zhang, Yuqing Song, Wei Wang WWW 2003 (3) Wei Wang, Jian Pei, Jiawei Han CIKM 2002 Haixun Wang, Wei Wang, Baile Shi, Peng Wang Yongtai Zhu, Wei Wang, Jian Pei, Baile Shi, Chen Wang KDD ICDM 2005 2004 (1) Wei Wang at UNC (3) Wei Wang at Fudan Univ., China (2) Wei Wang at UNSW, Australia (4) Wei Wang at SUNY Buffalo 118 The DISTINCT Methodology Measure similarity between references Link‐based similarity: Linkages between references References to the same object are more likely to be connected (Using random walk probability) Neighborhood similarity Neighbor tuples of each reference can indicate similarity between their contexts Self‐boosting: Training using the “same” bulky data set Reference‐based clustering Group references according to their similarities 119 Training with the “Same” Data Set Build a training set automatically Select distinct names, e.g., Johannes Gehrke The collaboration behavior within the same community share some similarity Training parameters using a typical and large set of “unambiguous” examples Use SVM to learn a model for combining different join paths Each join path is used as two attributes (with link‐based similarity and neighborhood similarity) The model is a weighted sum of all attributes 120 Clustering: Measure Similarity between Clusters Single‐link (highest similarity between points in two clusters) ? Complete‐link (minimum similarity between them)? No, because references to different objects can be connected. No, because references to the same object may be weakly connected. Average‐link (average similarity between points in two clusters)? A better measure Refinement: Average neighborhood similarity and collective random walk probability 121 Real Cases: DBLP Popular Names Name Num_authors Num_refs accuracy Hui Fang 3 9 1.0 1.0 1.0 1.0 Ajay Gupta 4 16 1.0 1.0 1.0 1.0 Joseph Hellerstein 2 151 0.81 1.0 0.81 0.895 Rakesh Kumar 2 36 1.0 1.0 1.0 1.0 Michael Wagner 5 29 0.395 1.0 0.395 0.566 Bing Liu 6 89 0.825 1.0 0.825 0.904 Jim Smith 3 19 0.829 0.888 0.926 0.906 Lei Wang 13 55 0.863 0.92 0.932 0.926 Wei Wang 14 141 0.716 0.855 0.814 0.834 Bin Yu 5 44 0.658 1.0 0.658 0.794 0.81 0.966 0.836 0.883 Average precision recall f-measure 122 Distinguishing Different “Wei Wang”s UNC-CH (57) Fudan U, China (31) Zhejiang U China (3) Najing Normal China (3) SUNY Binghamton (2) Ningbo Tech China (2) Purdue (2) Chongqing U China (2) Harbin U China (5) Beijing U Com China 5 2 UNSW, Australia (19) 6 SUNY Buffalo (5) Beijing Polytech (3) NU Singapore (5) (2) 123 Outline Motivation: Why Mining Heterogeneous Information Networks? Part I: Clustering, Ranking and Classification Clustering and Ranking in Information Networks Classification of Information Networks Part II: Data Quality and Search in Information Networks Data Cleaning and Data Validation by InfoNet Analysis Similarity Search in Information Networks Part III: Advanced Topics on Information Network Analysis Role Discovery and OLAP in Information Networks Mining Evolution and Dynamics of Information Networks Conclusions 124 Truth Validation by Info. Network Analysis The trustworthiness problem of the web (according to a survey): 54% of Internet users trust news web sites most of time 26% for web sites that sell products 12% for blogs TruthFinder: Truth discovery on the Web by link analysis Veracity (conformity to truth): Among multiple conflict results, can we automatically identify which one is likely the true fact? Given a large amount of conflicting information about many objects, provided by multiple web sites (or other information providers), how to discover the true fact about each object? Our work: Xiaoxin Yin, Jiawei Han, Philip S. Yu, “Truth Discovery with Multiple Conflicting Information Providers on the Web”, TKDE’08 125 Conflicting Information on the Web Different websites often provide conflicting info. on a subject, e.g., Authors of “Rapid Contextual Design” Online Store Authors Powell’s books Holtzblatt, Karen Barnes & Noble Karen Holtzblatt, Jessamyn Wendell, Shelley Wood A1 Books Karen Holtzblatt, Jessamyn Burns Wendell, Shelley Wood Cornwall books Holtzblatt-Karen, Wendell-Jessamyn Burns, Wood Mellon’s books Wendell, Jessamyn Lakeside books WENDELL, JESSAMYNHOLTZBLATT, KARENWOOD, SHELLEY Blackwell online Wendell, Jessamyn, Holtzblatt, Karen, Wood, Shelley 126 Our Setting: Info. Network Analysis Each object has a set of conflictive facts E.g., different author names for a book And each web site provides some facts How to find the true fact for each object? Web sites Facts w1 f1 w2 w3 w4 f2 Objects o1 f3 f4 o2 f5 127 Basic Heuristics for Problem Solving 1. 2. 3. 4. There is usually only one true fact for a property of an object This true fact appears to be the same or similar on different web sites E.g., “Jennifer Widom” vs. “J. Widom” The false facts on different web sites are less likely to be the same or similar False facts are often introduced by random factors A web site that provides mostly true facts for many objects will likely provide true facts for other objects 128 Overview of the TruthFinder Method Confidence of facts ↔ Trustworthiness of web sites A fact has high confidence if it is provided by (many) trustworthy web sites A web site is trustworthy if it provides many facts with high confidence The TruthFinder mechanism, an overview: Initially, each web site is equally trustworthy Based on the above four heuristics, infer fact confidence from web site trustworthiness, and then backwards Repeat until achieving stable state 129 Analogy to Authority‐Hub Analysis Facts ↔ Authorities, Web sites ↔ Hubs Web sites High trustworthiness w1 Hubs Facts f1 High confidence Authorities Difference from authority‐hub analysis Linear summation cannot be used A web site is trustable if it provides accurate facts, instead of many facts Confidence is the probability of being true Different facts of the same object influence each other 130 Inference on Trustworthness Inference of web site trustworthiness & fact confidence Web sites Facts w1 f1 w2 f2 w3 f3 w4 f4 Objects o1 o2 True facts and trustable web sites will become apparent after some iterations 131 Computational Model: t(w) and s(f) The trustworthiness of a web site w: t(w) Average confidence of facts it provides t (w) = ∑ f ∈F (w) s( f ) F (w ) Sum of fact confidence t(w1) w1 Set of facts provided by w The confidence of a fact f: s(f) One minus the probability that all web sites t(w2) providing f are wrong w2 s( f ) = 1 − ∏( ()1 − t (w)) s(f1) f1 Probability that w is wrong w∈W f Set of websites providing f 132 Experiments: Finding Truth of Facts Determining authors of books Dataset contains 1265 books listed on abebooks.com We analyze 100 random books (using book images) Case Voting TruthFinder Barnes & Noble Correct 71 85 64 Miss author(s) 12 2 4 Incomplete names 18 5 6 Wrong first/middle names 1 1 3 Has redundant names 0 2 23 Add incorrect names 1 5 5 No information 0 0 2 133 Experiments: Trustable Info Providers Finding trustworthy information sources Most trustworthy bookstores found by TruthFinder vs. Top ranked bookstores by Google (query “bookstore”) TruthFinder Bookstore trustworthiness #book Accuracy TheSaintBookstore 0.971 28 0.959 MildredsBooks 0.969 10 1.0 Alphacraze.com 0.968 13 0.947 Google rank #book Accuracy Barnes & Noble 1 97 0.865 Powell’s books 3 42 0.654 Google Bookstore 134 Outline Motivation: Why Mining Heterogeneous Information Networks? Part I: Clustering, Ranking and Classification Clustering and Ranking in Information Networks Classification of Information Networks Part II: Data Quality and Search in Information Networks Data Cleaning and Data Validation by InfoNet Analysis Similarity Search in Information Networks Part III: Advanced Topics on Information Network Analysis Role Discovery and OLAP in Information Networks Mining Evolution and Dynamics of Information Networks Conclusions 135 Similarity Search in Information Networks Structural similarity vs. semantic similarity Structural similarity: Based on structural/isomorphic similarity of sub‐graph/sub‐network structures Semantic similarity: influenced by similar network structures Graph‐structure‐based indexing and similarity search Structure‐based indexing, e.g., gIndex, Spath, … Use index to search for similar graph/network structures Substructure indexing Methods: Key problem: What substructures are good indexing features? gIndex [Yan, Yu & Han, SIGMOD’04]: Find frequent and discriminative subgraphs (by graph‐pattern mining) Spath [Zhao & Han, VLDB’10]: Use decomposed shortest paths as basic indexing features 136 Why S‐Path as Indexing Features? Shortest paths as neighborhood signatures of vertices (indexing features): scalable and pruning search space effectively Processing (by query decomposition): Decompose the query graph into a set of indexed shortest paths in SPath Query Network A global lookup table Neighborhood signature of v3 Semantics‐Based Similarity Search in InfoNet Search top‐k similar objects of the same type for a query Find researchers most similar with “Christos Faloutsos” Two critical concepts to define a similarity Feature space Traditional data: attributes denoted as numerical value/vector, set, etc. Network data: a relation sequence called “path schema” Existing homogeneous network‐based similarity does not deal with this problem Measure defined on the feature space Cosine, Euclidean distance, Jaccard coefficient, etc. PathSim 138 Path Schema for DBLP Queries Path schema: A path of InfoNet schema, e.g., APC, APA Who are most similar to Christos Faloutsos? 139 Flickr: Which Pictures Are Most Similar? Some path schema leads to similarity closer to human intuition But some others are not 140 Not All Similarity Measures Are Good Favor highly visible objects Not reasonable 141 Long Path Schema May Not Be Good Repeat the path schema 2, 4, and infinite times for conference similarity query 142 PathSim: Definition & Properties Commuting matrix corresponding to a path schema MP Product of adjacency matrix of relations in the path schema The element MP(i,j) denotes the strength between object i and object j in the semantic of path schema P If the weight of adjacency matrix is unweighted, it will denote the number of path instances following path schema P PathSim s(i,j) = 2MP(i,j)/(MP(i,i)+ MP(j,j)) Properties: 143 Co‐Clustering‐Based Pruning Algorithm Store commuting matrices for short path schemas & compute top‐k queries on line Framework Generate co‐clusters for materialized commuting matrices, for feature objects and target objects Derive upper bound for similarity between object and target cluster, and between object and object: Safely pruning target clusters and objects if the upper bound similarity is lower than current threshold Dynamically update top‐k threshold: Performance: Baseline vs. pruning 144 Outline Motivation: Why Mining Heterogeneous Information Networks? Part I: Clustering, Ranking and Classification Clustering and Ranking in Information Networks Classification of Information Networks Part II: Data Quality and Search in Information Networks Data Cleaning and Data Validation by InfoNet Analysis Similarity Search in Information Networks Part III: Advanced Topics on Information Network Analysis Role Discovery and OLAP in Information Networks Mining Evolution and Dynamics of Information Networks Conclusions 145 Role Discovery in Network: Why It Matters? Army communication network (imaginary) Automatically infer Commander Captain Solider 146 Role Discovery: Extraction Semantic Information from Links Objective: Extract semantic meaning from plain links to finely model and better organize information networks Challenges Latent semantic knowledge Interdependency Scalability Opportunity Human intuition Realistic constraint Crosscheck with collective intelligence Methodology: propagate simple intuitive rules and constraints over the whole network 147 Discovery of Advisor‐Advisee Relationships in DBLP Network Input: DBLP research publication network Output: Potential advising relationship and its ranking (r, [st, ed]) Ref. C. Wang, J. Han, et al., “Mining Advisor‐Advisee Relationships from Research Publication Networks”, SIGKDD 2010 Input: Temporal collaboration network Output: Relationship analysis 1999 Ada 2000 2 0 00 Visualized chorological hierarchies (0.9, [/, 1998]) Ada Bob (0.4, [/, 1998]) (0.5, [/, 2000]) 2000 (0.8, [1999,2000]) Jerry 2001 Ying 2002 Smith th (0.7, [2000, 2001]) 2004 Jerry Bob Ying 2003 (0.49, [/, 1999]) (0.2, [2001, 2003]) (0.65, [2002, 2004]) Smith 148 Overall Framework ai: author i pj: paper j py: paper year pn: paper# sti,yi: starting time edi,yi: ending time ri,yi: ranking score 149 Time‐Constrained Probabilistic Factor Graph (TPFG) yx: ax’s advisor stx,yx: starting time edx,yx: ending time • g(yx, stx, edx) is predefined local feature • fx(yx,Zx)= max g(yx , stx, edx) under time constraint • Objective function P({yx})=∏x fx (yx,Z) • Z={z| x ϵ Yz} • Yx: set of potential advisors of ax • • 150 Experiment Results DBLP data: 654, 628 authors, 1076,946 publications, years provided Labeled data: MathGealogy Project; AI Gealogy Project; Homepage Datasets RULE SVM IndMAX TEST1 69.9% 73.4% 75.2% 78.9% 80.2% 84.4% TEST2 69.8% 74.6% 74.6% 79.0% 81.5% 84.3% TEST3 80.6% 86.7% 83.1% 90.9% 88.8% 91.3% heuristics Supervised learning TPFG Empirical optimized parameter parameter 151 Case Study & Scalability Advisee Time Note David M. 1. Michael I. Jordan Blei 2. John D. Lafferty 01‐03 PhD advisor, 2004 grad 05‐06 Postdoc, 2006 Hong Cheng 1. Qiang Yang 02‐03 MS advisor, 2003 2. Jiawei Han 04‐08 PhD advisor, 2008 1. Rajeev Motawani 97‐98 “Unofficial advisor” Sergey Brin Top Ranked Advisor 152 Graph/Network Summarization: Graph Compression Extract common subgraphs and simplify graphs by condensing these subgraphs into nodes 153 OLAP on Information Networks Why OLAP information networks? Advantages of OLAP: Interactive exploration of multi‐dimensional and multi‐level space in a data cube Infonet Multi‐dimensional: Different perspectives Multi‐level: Different granularities InfoNet OLAP: Roll‐up/drill‐down and slice/dice on information network data Traditional OLAP cannot handle this, because they ignore links among data objects Handling two kinds of InfoNet OLAP Informational OLAP Topological OLAP 154 Informational OLAP In the DBLP network, study the collaboration patterns among researchers Dimensions come from informational attributes attached at the whole snapshot level, so‐called Info‐Dims I‐OLAP Characteristics: Overlay multiple pieces of information No change on the objects whose interactions are being examined In the underlying snapshots, each node is a researcher In the summarized view, each node is still a researcher 155 Topological OLAP Dimensions come from the node/edge attributes inside individual networks, so‐called Topo‐Dims T‐OLAP Characteristics Zoom in/Zoom out Network topology changed: “generalized” nodes and “generalized” edges In the underlying network, each node is a researcher In the summarized view, each node becomes an institute that comprises multiple researchers 156 InfoNet OLAP: Operations & Framework InfoNet I‐OLAP InfoNet T‐OLAP Overlay multiple snapshots to Shrink the topology & obtain a T‐aggregated form a higher‐level summary via network that represents a compressed view, I‐aggregated network with topological elements (i.e., nodes and/or Roll‐up edges) merged and replaced by corresp. higher‐level ones Return to the set of lower‐level A reverse operation of roll‐up Drill‐down snapshots from the higher‐level overlaid (aggregated) network Select a subset of qualifying Select a subnetwork based on Topo‐Dims Slice/dice snapshots based on Info‐Dims Measure is an aggregated graph & other measures like node count, average degree, etc. can be treated as derived Graph plays a dual role: (1) data source, and (2) aggregate measure Measures could be complex, e.g., maximum flow, shortest path, centrality It is possible to combine I‐OLAP and T‐OLAP into a hybrid case 157 Outline Motivation: Why Mining Heterogeneous Information Networks? Part I: Clustering, Ranking and Classification Clustering and Ranking in Information Networks Classification of Information Networks Part II: Data Quality and Search in Information Networks Data Cleaning and Data Validation by InfoNet Analysis Similarity Search in Information Networks Part III: Advanced Topics on Information Network Analysis Role Discovery and OLAP in Information Networks Mining Evolution and Dynamics of Information Networks Conclusions 158 Mining Evolution and Dynamics of InfoNet Many networks are with time information E.g., according to paper publication year, DBLP networks can form network sequences Motivation: Model evolution of communities in heterogeneous network Automatically detect the best number of communities in each timestamp Model the smoothness between communities of adjacent timestamps Model the evolution structure explicitly Birth, death, split 159 Evolution: Idea Illustration From network sequences to evolutionary communities 160 Graphical Model: A Generative Model Dirichlet Process Mixture Model‐based generative model At each timestamp, a community is dependent on historical communities and background community distribution 161 Generative Model & Model Inference To generate a new paper oi Decide whether to join an existing community or a new one Join an existing community k with prob. nk/(i‐1+α) Join a new community k with prob. α/(i‐ 1 +α): Decide its prior, either from a background distribution (λ) or historical communities ((1‐ λ)πk), with different probabilities, draw the attribute distribution from the prior Generate oi according to the attribute distribution Greedy inference for each timestamp: Collapse Gibbs sampling, which is trying to sample cluster label for each target object (e.g., paper) 162 Accuracy Study The more types of objects used, the better accuracy Historical prior results in better accuracy 163 Case Study on DBLP Tracking database community evolution 1991 1994 1997 2000 DEXA DEXA DEXA Workshop VLDB Comm, ACM VLDB SIGMOD Conf. ICDE ICDE SIGMOD Conf. VLDB SIGMOD Conf. VLDB CIKM DEXA DEXA Int. J. MM Studies TKDE ICDE IDEAS systems object data data object oriented databases databases database database database database information systems object query oriented databases systems web (1993) DBLP Schema CHI Conf. Comp. TREC AAAI SIGIR Comm. ACM PKDD W. Simu. Conf. SIGKDD Explor. SIGCSE KDD software data systems retrieval design information knowledge mining analysis text 164 Case Study on Delicious.com Delicious Schema 550 C1 500 C2 Event Count 450 C3 400 350 300 250 200 150 1 1.5 2 2.5 Week 3 3.5 4 165 Outline Motivation: Why Mining Heterogeneous Information Networks? Part I: Clustering, Ranking and Classification Clustering and Ranking in Information Networks Classification of Information Networks Part II: Data Quality and Search in Information Networks Part III: Advanced Topics on Information Network Analysis Role Discovery and OLAP in Information Networks Mining Evolution and Dynamics of Information Networks Conclusions Data Cleaning and Data Validation by InfoNet Analysis Similarity Search in Information Networks Conclusions Rich knowledge can be mined from information networks What is the magic? Heterogeneous, structured information networks! Clustering, ranking and classification: Integrated clustering, ranking and classification: RankClus, NetClus, GNetMine, … Data cleaning, validation, and similarity search Role discovery, OLAP, and evolutionary analysis Knowledge is power, but knowledge is hidden in massive links! Mining heterogeneous information networks: Much more to be explored!! 167 Future Research From mining current single star network schema to ranking, clustering, …, in multi‐star, multi‐relational databases Mining information networks formed by structured data linking with unstructured data (text, multimedia and Web) Mining cyber‐physical networks (networks formed by dynamic sensors, image/video cameras, with information networks) Enhancing the power of knowledge discovery by transforming massive unstructured data: Incremental information extraction, role discovery, … ⇒ multi‐dimensional structured info‐net Mining noisy, uncertain, un‐trustable massive datasets by information network analysis approach Turning Wikipedia and/or Web into structured or semi‐structured databases by heterogeneous information network analysis 168 References: Books on Network Analysis A.‐L. Barabasi. Linked: How Everything Is Connected to Everything Else and What It Means. Plume, 2003. M. Buchanan. Nexus: Small Worlds and the Groundbreaking Theory of Networks. W. W. Norton & Company, 2003. D. J. Cook and L. B. Holder. Mining Graph Data. John Wiley & Sons, 2007 S. Chakrabarti. Mining the Web: Discovering Knowledge from Hypertext Data. Morgan Kaufmann, 2003 A. Degenne and M. Forse. Introducing Social Networks. Sage Publications, 1999 P. J. Carrington, J. Scott, and S. Wasserman. Models and Methods in Social Network Analysis. Cambridge University Press, 2005. J. Davies, D. Fensel, and F. van Harmelen. Towards the Semantic Web: Ontology‐Driven Knowledge Management. John Wiley & Sons, 2003. D. Fensel, W. Wahlster, H. Lieberman, and J. Hendler. Spinning the Semantic Web: Bringing the World Wide Web to Its Full Potential. MIT Press, 2002. L. Getoor and B. Taskar (eds.). Introduction to statistical learning. In MIT Press, 2007. B. Liu. Web Data Mining: Exploring Hyperlinks, Contents, and Usage Data. Springer, 2006. J. P. Scott. Social Network Analysis: A Handbook. Sage Publications, 2005. J. Watts. Six Degrees: The Science of a Connected Age. W. W. Norton & Company, 2003. D. J.Watts. Small Worlds: The Dynamics of Networks between Order and Randomness. Princeton University Press, 2003. S. Wasserman and K. Faust. Social Network Analysis: Methods and Applications. Cambridge University Press, 1994. References: Some Overview Papers T. Berners‐Lee, J. Hendler, and O. Lassila. The semantic web. Scientific American, May 2001. C. Cooper and A Frieze. A general model of web graphs. Algorithms, 22, 2003. S. Chakrabarti and C. Faloutsos. Graph mining: Laws, generators, and algorithms. ACM Comput. Surv., 38, 2006. T. Dietterich, P. Domingos, L. Getoor, S. Muggleton, and P. Tadepalli. Structured machine learning: The next ten years. Machine Learning, 73, 2008 S. Dumais and H. Chen. Hierarchical classification of web content. SIGIR'00. S. Dzeroski. Multirelational data mining: An introduction. ACM SIGKDD Explorations, July 2003. L. Getoor. Link mining: a new data mining challenge. SIGKDD Explorations, 5:84{89, 2003. L. Getoor, N. Friedman, D. Koller, and B. Taskar. Learning probabilistic models of relational structure. ICML'01 D. Jensen and J. Neville. Data mining in networks. In Papers of the Symp. Dynamic Social Network Modeling and Analysis, National Academy Press, 2002. T. Washio and H. Motoda. State of the art of graph‐based data mining. SIGKDD Explorations, 5, 2003. 170 References: Some Influential Papers A. Z. Broder, R. Kumar, F. Maghoul, P. Raghavan, S. Rajagopalan, R. Stata, A. Tomkins, and J. L. Wiener. Graph structure in the web. Computer Networks, 33, 2000. S. Brin and L. Page. The anatomy of a large‐scale hyper‐textual web search engine. WWW'98. S. Chakrabarti, B. E. Dom, S. R. Kumar, P. Raghavan, S. Rajagopalan, A. Tomkins, D. Gibson, and J. M. Kleinberg. Mining the web's link structure. COMPUTER, 32, 1999. M. Faloutsos, P. Faloutsos, and C. Faloutsos. On power‐law relationships of the internet topology. ACM SIGCOMM'99 M. Girvan and M. E. J. Newman. Community structure in social and biological networks. In Proc. Natl. Acad. Sci. USA 99, 2002. B. A. Huberman and L. A. Adamic. Growth dynamics of world‐wide web. Nature, 399:131, 1999. G. Jeh and J. Widom. SimRank: a measure of structural‐context similarity. KDD'02 J. M. Kleinberg, R. Kumar, P. Raghavan, S. Rajagopalan, and A. Tomkins. The web as a graph: Measurements, models, and methods. COCOON'99 D. Kempe, J. Kleinberg, and E. Tardos. Maximizing the spread of influence through a social network. KDD'03 J. M. Kleinberg. Small world phenomena and the dynamics of information. NIPS'01 R. Kumar, P. Raghavan, S. Rajagopalan, D. Sivakumar, A. Tomkins, and E. Upfal. Stochastic models for the web graph. FOCS'00 M. E. J. Newman. The structure and function of complex networks. SIAM Review, 45, 2003. 171 References: Clustering and Ranking (1) E. Airoldi, D. Blei, S. Fienberg and E. Xing, “Mixed Membership Stochastic Blockmodels”, JMLR’08 Liangliang Cao, Andrey Del Pozo, Xin Jin, Jiebo Luo, Jiawei Han, and Thomas S. Huang, “RankCompete: Simultaneous Ranking and Clustering of Web Photos”, WWW’10 G. Jeh and J. Widom, “SimRank: a measure of structural‐context similarity”, KDD'02 Jing Gao, Feng Liang,Wei Fan, Chi Wang, Yizhou Sun, and Jiawei Han, “Community Outliers and their Efficient Detection in Information Networks", KDD'10 M. E. J. Newman and M. Girvan, “Finding and evaluating community structure in networks”, Physical Review E, 2004 M. E. J. Newman and M. Girvan, “Fast algorithm for detecting community structure in networks”, Physical Review E, 2004 J. Shi and J. Malik, “Normalized cuts and image Segmentation”, CVPR'97 Yizhou Sun, Yintao Yu, and Jiawei Han, "Ranking‐Based Clustering of Heterogeneous Information Networks with Star Network Schema", KDD’09 Yizhou Sun, Jiawei Han, Peixiang Zhao, Zhijun Yin, Hong Cheng, and Tianyi Wu, "RankClus: Integrating Clustering with Ranking for Heterogeneous Information Network Analysis", EDBT’09 172 References: Clustering and Ranking (2) Yizhou Sun, Jiawei Han, Jing Gao, and Yintao Yu, "iTopicModel: Information Network‐ Integrated Topic Modeling", ICDM’09 Xiaoxin Yin, Jiawei Han, Philip S. Yu. "LinkClus: Efficient Clustering via Heterogeneous Semantic Links", VLDB'06. Yintao Yu, Cindy X. Lin, Yizhou Sun, Chen Chen, Jiawei Han, Binbin Liao, Tianyi Wu, ChengXiang Zhai, Duo Zhang, and Bo Zhao, "iNextCube: Information Network‐ Enhanced Text Cube", VLDB'09 (demo) A. Wu, M. Garland, and J. Han. Mining scale‐free networks using geodesic clustering. KDD'04 Z. Wu and R. Leahy, “An optimal graph theoretic approach to data clustering: Theory and its application to image segmentation”, IEEE Trans. Pattern Anal. Mach. Intell., 1993. X. Xu, N. Yuruk, Z. Feng, and T. A. J. Schweiger. SCAN: A structural clustering algorithm for networks. KDD'07 X. Yin, J. Han, and P. S. Yu. Cross‐relational clustering with user's guidance. KDD'05 173 References: Network Classification (1) A. Appice, M. Ceci, and D. Malerba. Mining model trees: A multi‐relational approach. ILP'03 Jing Gao, Feng Liang, Wei Fan, Yizhou Sun, and Jiawei Han, "Bipartite Graph‐based Consensus Maximization among Supervised and Unsupervised Models ", NIPS'09 L. Getoor, N. Friedman, D. Koller and B. Taskar, “Learning Probabilistic Models of Link Structure”, JMLR’02. L. Getoor, E. Segal, B. Taskar and D. Koller, “Probabilistic Models of Text and Link Structure for Hypertext Classification”, IJCAI WS ‘Text Learning: Beyond Classification’, 2001. L. Getoor, N. Friedman, D. Koller, and A. Pfeffer, “Learning Probabilistic Relational Models”, chapter in Relation Data Mining, eds. S. Dzeroski and N. Lavrac, 2001. M. Ji, Y. Sun, M. Danilevsky, J. Han, and J. Gao. “Graph‐based classification on heterogeneous information networks”, ECMLPKDD’10. Q. Lu and L. Getoor, “Link‐based classification”, ICML'03 D. Liben‐Nowell and J. Kleinberg, “The link prediction problem for social networks”, CIKM'03 174 References: Network Classification (2) J. Neville, B. Gallaher, and T. Eliassi‐Rad. Evaluating statistical tests for within‐network classifiers of relational data. ICDM'09. J. Neville, D. Jensen, L. Friedland, and M. Hay. Learning relational probability trees. KDD'03 Jennifer Neville, David Jensen, “Relational Dependency Networks”, JMLR’07 M. Szummer and T. Jaakkola, “Partially labeled classication with markov random walks”, In NIPS, volume 14, 2001. M. J. Rattigan, M. Maier, and D. Jensen. Graph clustering with network structure indices. ICML'07 P. Sen, G. M. Namata, M. Galileo, M. Bilgic, L. Getoor, B. Gallagher, and T. Eliassi‐Rad. Collective classification in network data. AI Magazine, 29, 2008. B. Taskar, E. Segal, and D. Koller. Probabilistic classification and clustering in relational data. IJCAI'01 B. Taskar, P. Abbeel, M.F. Wong, and D. Koller, “Relational Markov Networks”, chapter in L. Getoor and B. Taskar, editors, Introduction to Statistical Relational Learning, 2007 X. Yin, J. Han, J. Yang, and P. S. Yu, “CrossMine: Efficient Classification across Multiple Database Relations”, ICDE'04. D. Zhou, O. Bousquet, T. N. Lal, J. Weston, and B. Scholkopf, “Learning with local and global consistency”, In NIPS 16, Vancouver, Canada, 2004. X. Zhu and Z. Ghahramani, “Learning from labeled and unlabeled data with label propagation”, Technical Report, 2002. 175 References: Social Network Analysis B. Aleman‐Meza, M. Nagarajan, C. Ramakrishnan, L. Ding, P. Kolari, A. P. Sheth, I. B. Arpinar, A. Joshi, and T. Finin. Semantic analytics on social networks: experiences in addressing the problem of conflict of interest detection. WWW'06 R. Agrawal, S. Rajagopalan, R. Srikant, and Y. Xu. Mining newsgroups using networks arising from social behavior. WWW'03 P. Boldi and S. Vigna. The WebGraph framework I: Compression techniques. WWW'04 D. Cai, Z. Shao, X. He, X. Yan, and J. Han. Community mining from multi‐relational networks. PKDD'05 P. Domingos. Mining social networks for viral marketing. IEEE Intelligent Systems, 20, 2005. P. Domingos and M. Richardson. Mining the network value of customers. KDD'01 P. DeRose, W. Shen, F. Chen, A. Doan, and R. Ramakrishnan. Building structured web community portals: A top‐down, compositional, and incremental approach. VLDB'07 G. Flake, S. Lawrence, C. L. Giles, and F. Coetzee. Self‐organization and identification of web communities. IEEE Computer, 35, 2002. J. Kubica, A. Moore, and J. Schneider. Tractable group detection on large link data sets. ICDM'03 176 References: Data Quality & Search in Networks I. Bhattacharya and L. Getoor, “Iterative record linkage for cleaning and integration”, Proc. SIGMOD 2004 Workshop on Research Issues on Data Mining and Knowledge Discovery (DMKD'04) Xin Luna Dong, Laure Berti‐Equille, and Divesh Srivastava, “Integrating conflicting data: The role of source dependence”, PVLDB, 2(1):550–561, 2009. Xin Luna Dong, Laure Berti‐Equille, and Divesh Srivastava, “Truth discovery and copying detection in a dynamic world”, PVLDB, 2(1):562–573, 2009. H. Han, L. Giles, H. Zha, C. Li, and K. Tsioutsiouliklis, “Two supervised learning approaches for name disambiguation in author citations”, ICDL'04. Y. Sun, J. Han, T. Wu, X. Yan, and Philip S. Yu, “PathSim: Path Schema‐Based Top‐K Similarity Search in Heterogeneous Information Networks”, Technical report, CS, UIUC, July 2010. X. Yin, J. Han, and P. S. Yu, “Object Distinction: Distinguishing Objects with Identical Names by Link Analysis”, ICDE'07 X. Yin, J. Han, and P. S. Yu, “Truth Discovery with Multiple Conflicting Information Providers on the Web”, IEEE TKDE, 20(6):796‐808, 2008 P. Zhao and J. Han, “On Graph Query Optimization in Large Networks”, VLDB’10. 177 References: Role Discovery, Summarization and OLAP D. Archambault, T. Munzner, and D. Auber. Topolayout: Multilevel graph layout by topological features. IEEE Trans. Vis. Comput. Graph, 2007. Chen Chen, Xifeng Yan, Feida Zhu, Jiawei Han, and Philip S. Yu, "Graph OLAP: Towards Online Analytical Processing on Graphs", ICDM 2008 Chen Chen, Xifeng Yan, Feida Zhu, Jiawei Han, and Philip S. Yu, "Graph OLAP: A Multi‐ Dimensional Framework for Graph Data Analysis", KAIS 2009. Xin Jin, Jiebo Luo, Jie Yu, Gang Wang, Dhiraj Joshi, and Jiawei Han, “iRIN: Image Retrieval in Image Rich Information Networks”, WWW'10 (demo paper) Lu Liu, Feida Zhu, Chen Chen, Xifeng Yan, Jiawei Han, Philip Yu, and Shiqiang Yang, “Mining Diversity on Networks", DASFAA'10 Y. Tian, R. A. Hankins, and J. M. Patel. Efficient aggregation for graph summarization. SIGMOD'08 Chi Wang, Jiawei Han, Yuntao Jia, Jie Tang, Duo Zhang, Yintao Yu, and Jingyi Guo, “Mining Advisor‐Advisee Relationships from Research Publication Networks ", KDD'10 Zhijun Yin, Manish Gupta, Tim Weninger and Jiawei Han, “LINKREC: A Unified Framework for Link Recommendation with User Attributes and Graph Structure ”, WWW’10 178 References: Network Evolution L. Backstrom, D. Huttenlocher, J. Kleinberg, and X. Lan. Group formation in large social networks: Membership, growth, and evolution. KDD'06 M.‐S. Kim and J. Han. A particle‐and‐density based evolutionary clustering method for dynamic networks. VLDB'09 J. Leskovec, J. Kleinberg, and C. Faloutsos. Graphs over time: Densification laws, shrinking diameters and possible explanations. KDD'05 Yizhou Sun, Jie Tang, Jiawei Han, Manish Gupta, Bo Zhao, “Community Evolution Detection in Dynamic Heterogeneous Information Networks”, KDD‐ MLG’10 179